Large Language Models (LLMs) have been shown to aid human experts in the development of Assurance Cases (ACs), including supporting tasks like defeater generation and generating AC argument structures. As a result, LLMs have the potential to both improve the quality of ACs and reduce the cost of their development. However, ACs are often prepared for high-risk systems and applications, and the consequences of poor judgement or errors in the assurance process can be severe. Therefore, it is important to consider both the potential risks and benefits of using LLMs to support AC development. To this end, the contribution of this paper is four-fold. First, the paper surveys the literature on the use of LLMs to support ACs and identifies four LLM use cases. Second, this paper suggests seven additional novel LLM use cases. Third, this paper proposes a method for assessing the risk-benefit trade-off of a specific LLM use case. Finally, the paper applies the proposed method to the identified use cases, which demonstrates the applicability of the method and provides an opportunity to evaluate the benefits and consequences associated with each use case.